Over the past week or so, there have been several advertisements for
free software packages for dealing with Bayesian Networks, none of which
are mentioned in Russ Almond's list of BN software
(http://bayes.stat.washington.edu/almond/belief.html).
So I have made my own list:
http://HTTP.CS.Berkeley.EDU/~murphyk/bnsoft.html.
If you would like your software mentioned on this list, please send me
email. Also, if I have made an error in describing the features of your
program, I apologize, and will correct it if you send me email.
(I have not actually tried many of these programs, so the feature list
is based on reading the web pages.)
In keeping with this trend, I am also announcing my own software
package, the Bayes Net Toolbox, written in Matlab 5.
The main features that distinguish it from other packages are: it
supports discrete and continuous (Gaussian) nodes, static and dynamic
nets (including HMMs and Kalman filters), parameter and structure
learning, and it uses several different inference algorithms (including
join tree, Gaussian marginalization, forwards-backwards (in C) and
Kalman-RTS). See http://www.cs.berkeley.edu/~murphyk/bnt.html if you
are interested.
Kevin